Update: DeepPavlov Community Call #5 RecordingSubscribe for future calls here (last Thursday of the month, 8am Pacific/7pm MSK): http://bit.ly/MonthlyDPCommunityCall2021Dear DeepPavlov community,The online DeepPavlov Community Call is scheduled for January 28th!This is the first call in 2021, and it will be all about our plans for DeepPavlov Library Refactoring Project. It is a big project for us, and we can’t wait to share our news and hear your thoughts.As always, you are welcome to suggest topics and call in!
DeepPavlov Community Call #5 (January 28, 2021)
We’ll hold the fifth one on January 28, 2021 at 8:00am Pacific (7pm MSK/4 or 5pm UTC depending on DST).
Add to your calendar:
Last two years were fantastic for us. Participation in famous challenges like Alexa Prize 3, Alexa Prize 4 as well as Russian Pro//Chtenie helped us to battle test our DeepPavlov Library in the real conditions. It was fascinating for us to see how our Library fueled our endeavors in the Conversational AI space!
DeepPavlov lab was created to enable building super smart, intelligent multiskill AI Assistants. To make this happen, we’ve created 3 key components:
- DeepPavlov Library with a huge (200+) number of different configs that include NLP models that span across basic NLP tasks like entity extraction, sentiment analysis, classification; frameworks for building complex skills like Go-Bot and FAQ; integration with different systems; and myriads of examples. First release was shipped on February 5, 2018.
- DeepPavlov Agent which is our Conversational AI Orchestrator. It enabled our efforts on Alexa Prize 3, DeepPavlov Dream AI Assistant Demo, Deepy, and is currently the heart of our Alexa Prize 4 Dream socialbot. It’s also versatile enough to support solutions for sister NLP problems like language comprehension that helped us to win at one of Pro//Chtenie challenges. First release was shipped on June 28, 2019.
- DeepPavlov Deepy which is our platform for building Multiskill AI Assistants. It was born from our experimental DeepPavlov Dream AI Assistant Demo which itself was born from our Alexa Prize 3 Dream socialbot. Deepy provides a number of different goal-oriented, FAQ-based, and chit-chat skills, and several examples of Distributions that showcase different kinds of multiskill AI assistants you can build with DeepPavlov technology. First release was shipped on December 8, 2020.
Releases of Agent and Deepy allowed us to round up our efforts in helping you to build an end-to-end solution. However, while some of the components of the original Library were efficiently integrated with Agent and Deepy, some remained in their experimental form. Others became effectively obsolete.
It’s now time to begin our Library’s transformation on its path to v1.0.
This process will involve a lot of restructuring, refactoring that will in turn lead to significant changes. Here are some of the areas:
These are some of the top areas that we will focus in the coming months.
- Migration to PyTorch
- Deprecation of Old Models & Code
- Models Version Control
- Updates to Configs
We welcome you to join us at our DeepPavlov Community Call #5 to let us know what you think about these changes, share your expectations from v1 of the Library, and tell us how DeepPavlov Library helps you in your projects!
Agenda for the DeepPavlov Community Call #5:
- 7:00pm–7:30pm | Welcome and overview of DeepPavlov Library changes in 2021H1
- 7:30pm–8:00pm | Time for Feedback & Q&A with the team
In case you’ve missed the fourth one, we’ve uploaded a record — see the playlist. Check it out!
DeepPavlov Library Feedback Survey
We want to hear your thoughts. Starting with today, you can fill in this form to let us know how you use DeepPavlov Library, what you want us to add or improve!
We are anxious to hear your thoughts!
Please let us know and leave a comment with any topics or questions you’d like to hear about!
We can’t promise to cover everything but we’ll do our best next week or in a future call.
After calling in or watching, please do fill in the survey to let us know if you have any feedback on the format or content: https://bit.ly/dpcallsurvey
The DeepPavlov team